Вы находитесь на странице: 1из 22

225

CFD Prediction of Erosion Wear in


Centrifugal Slurry Pumps for Dilute
Slurry Flows
K. V. Pagalthivarthi1, P. K. Gupta2*, Vipin Tyagi3, M. R. Ravi4
1CFD Research Leader, GIW Industries Inc., Grovetown, GA 30813, USA
2*Professor.,Dept. of Mech. Engg., Raghu Engg. College, Visakhapatnam 531162
3Research Analyst, SHELL Technology India, Bangalore 560066, India
4Professor, Dept. of Applied Mechanics, IIT Delhi, New Delhi - 110016, India

Received: 25 April 2011, Accepted: 10 October 2011

Abstract
The paper discusses numerical prediction of erosion wear trends in centrifugal pump
casing pumping dilute slurries. The casing geometry is considered two-dimensional.
Discrete Phase Model (DPM) in FLUENT 6.1 is utilized to obtain dilute slurry flow
field through the pump casing employing two-way coupling. Standard k  e model is
used for turbulence. Effect of several operational parameters viz. pump flow rate, pump
speed (RPM), particle diameter and various geometry conditions viz. tongue curvature,
slope of the discharge pipe and casing width is studied. Qualitative trends of erosion
wear is described for these operational and geometric parameters with an idea to lower
the wear rates and to make the wear pattern along the casing wall as uniform as
possible. For example, with increase in pump flow rate, wear rates tends to even out
whereas with increased casing width, wear rates are found to decrease.

1. INTRODUCTION
Centrifugal pumps are widely used in transporting solid-liquid mixtures (slurries) [1] with
applications in mining industry, metallurgical operations, chemical industry, coal industries and so
on. Performance reduction of the pump [2] and erosion damage of the wetted components are two
important and frequently faced problems when pumping solid-liquid mixtures. The important pump
wetted components that undergo severe erosion effects are the liners, the impeller and the pump
casing. The pump casing is the heaviest component among the three wearing parts and its design is
also very complex. Hence it is the costliest wearing component of the pump.
In abrasive slurry applications, hard cast iron alloys (usually chromium-nickel) are used for the
wetted components. Due to the difficulty in repairing such hard alloy cast iron materials, a local
damaged area could render an otherwise serviceable component to be rejected. This leads to
frequent replacement of spare parts. Each time a part is to be replaced, a complete shutdown of the
pipeline operation is also necessitated. The net effect of this is that the overall cost of ownership
[3] of the slurry pumps escalates. For example, it has been reported [4] that an unscheduled
downtime can cost in excess of 100,000 US$ per hour for a SAG mill slurry pump operating under
the most severe of wear conditions. Therefore, from the viewpoints of useful life of pump
components, cost of repairing/replacing worn out parts, and the associated unscheduled downtime,
erosive wear prediction has attracted considerable interest and concern over the years.
Erosion in slurry pumps occurs due to two mechanisms, viz., particle impact and sliding
(scouring) action [5,6,7]. In dilute slurries, erosion is likely to be more due to particle impact [8].
Dilute slurry flow in slurry transportation is of significance in handling sewage, silted waters and
certain dredging and mining applications. In such situations, the absence of a protective sliding bed
combined with appreciable deviations in the trajectories of the particles from the streamlines (Fig.
1) of the carrier phase facilitates wear by particle impact (known as impact wear). Particle impact
could also be directional or random (due to turbulence).
Erosion wear is calculated via wear mechanisms in which the flow properties of the slurry, slurry
material and casing material are related together with certain empirical constants. Thus, erosion
226 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows

Carrier Phase flow cle


rti
Near the wall Pa

Figure 1. Particle-impact erosion due to the deviation of the particle


from the streamlines.

wear prediction comprises of two steps viz. (1) to compute the two-phase flow field in the relevant
component (impeller, liner and casing), and (2) relating the computed flow field to the local wear
rate via experimentally determined wear coefficients. Experimental procedures for measuring both
impact and sliding wear coefficients are described elsewhere [5,9,10,11,12] in the literature.
Although in principle, a 3-D modeling of the full pump could be considered to obtain accurate
details of the flow field, this approach, however, will prove very expensive due to complex pump
geometries, rotating impeller and other modeling complications. Alternately, we could also
consider a component-by-component analysis of the pump provided that the appropriate boundary
conditions are specified for each component. A component-by-component approach serves readily
the specific needs of a design engineer [13] and is also computationally efficient. Thus, this
approach has been used in the present study to evaluate erosion wear in centrifugal pump casing by
applying suitable boundary conditions.
Two dimensional models of the pump casings have been developed in the past [7, 14, 15]. These
studies [7, 14, 15] included an inviscid approach in modeling slurry flow and had some crucial
limitations viz., neglecting the viscous terms for particles greater than 1mm diameter, the procedure
did not converge reliably. The models in these studies [7, 14, 15] allowed some mass loss of the
carrier phase flow through the casing wall. Neglect of pressure and gravity terms in force balance
on the particles also limited the analysis to fine particles which follow the carrier fluid closely.
These limitations were addressed in [16] using a quasi-3D model. The main contribution of the
study [16] was the prediction of main stream and secondary flows at any arbitrary cross-section of
the pump casing.
Zhong and Minemura [17] applied potential formulation for solving the carrier phase equations.
Particle velocities used were determined using Lagrangian treatment by balancing forces. However,
they used a different erosion model, viz. Bitters erosion model [18], for determining erosion along
the casing wall. Later Addie et al. [19] compared the viscous (RANS) finite element computations
with the Particle Image Velocimetry (PIV) results [20] and found reasonable agreement within the
scope of the modeling assumptions. Various other studies [21] have addressed the influence of flow
and geometry parameters on the velocity and pressure distribution in centrifugal pump casing using
CFD.
Wear studies specific to pump casings have been performed by [14, 15, 22, 7, 9, 23, 17]. In these
studies, wear rate pattern along the casing wall was obtained at various operating conditions and
for various particle diameters and inlet concentrations.
In the present work, the focus is on qualitative study of erosion wear in centrifugal slurry pump
casing for dilute slurry flow. For dilute slurries, the hybrid Eulerian-Lagrangian approach is most
effective [24] in flow prediction. Zhong and Minemura [17] used LagrangianEulerian approach
for predicting slurry flow inside centrifugal pump casing. Lagrangian models were also developed
by many researchers for prediction of erosion in pipes and elbows such as [25, 26, 27, 28]. These
studies considered only one-way coupling between the phases. In one-way coupling solid particles
do not interact with each other and particle motion does not affect the fluid-flow field.
In the present study using FLUENT 6.1, the slurry is modeled using a viscous model with two-
way coupling. Two way coupling accounts for the influence of particulate motion on the carrier-
phase and vice-versa. As in previous studies, the aim is to obtain the wear rate along the casing
periphery. The aim of wear modeling is to help select a pump (for a specific application) such that
a nearly even wear pattern is obtained without significant local gouging effects. Even if such a
selection of pump is not possible, a prediction of the expected wear pattern can still be useful in
providing extra material thickness in regions of heavy wear. As aforementioned, numerical

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 227

modeling is important in understanding, predicting and controlling the wear process. The solid-
liquid flow properties adjacent to the wear surface are first computed using FLUENT 6.1. Erosion
wear is then computed using particle impact wear assumptions.
Particle trajectories and their distribution along the casing wall have been presented for various
operating parameters such as pump speed (RPM), flow rate, casing geometry and particle size.
Erosion of pump casing wall due to variations in velocity and concentration are qualitatively
presented.

2. MATHEMATICAL FORMULATION
Discrete Phase Model (DPM) in FLUENT 6.1 has been used to compute dilute slurry flow field.
Two-way coupling between the solid and liquid phases has been considered, which implies that the
carrier phase affects the particles and vice versa via drag term. The particle trajectory equations
and carrier phase equations are solved in a coupled manner. The coupling is affected by adding a
source term in the momentum equation of the carrier phase as obtained from momentum exchange
between the phases.
The momentum exchange by the particle is given as

F=
particles
Fdrag m p t (1)

where Fdrag is the drag force .per unit mass of the particles, t is the time step used for integrating
the trajectory equation and m p is the mass flow rate of particles. Other forces like Saffmans lift
force can also be included in the exchange term with the drag force. This momentum exchange
appears as a momentum sink in the continuous phase momentum balance in any subsequent
calculation of the continuous phase flow field. The modified carrier phase equations are then given
as (in non-dimensional form)

u u 2(1 + r ) u (1 + r ) u v
u +v = p + + Fx (2)
x y x Re x y Re y x

v v (1 + r ) u v 2(1 + r ) v
u +v = + + p Fy (3)
x y x Re y x y Re y

where Fx and Fy are non-dimensionalized components of the momentum exchange term. In Eqns.
(23), u and v are carrier-phase velocity components non-dimensionalized w.r.t impeller tip
velocity U0; p is pressure non-dimensionalized with respect to rLU2O where rL is the carrier-phase
density; mr  mL  mt, mL and mt being respectively, the laminar and turbulent viscosities of the
carrier-phase; and Re is the carrier-phase flow Reynolds number.
Particle trajectories are predicted by integrating the force balance on the particles, which is
written in a Lagrangian reference frame. The force balance on a particle equating inertia with forces
is given as

du p  
(  p
= FD u u p + g )

+F
( ) (4)
dt p

where u p , the particle velocity, is non dimensionalized with impeller tip velocity, U0, time is
non-dimensionalized by R/U0 and gravity ( g~ ) is non-dimensionalized as gR/U20, R being the
radius of the impeller. The first term on the right-hand side of Equation (4) is the force due to
relative motion of the particle in the fluid, i.e. the drag force. The second term describes the
combined effect of gravity and buoyancy, while the third term describes the additional forces viz.
lift force and virtual mass force per unit mass of particles. Saffman lift force is intended for small
particle Reynolds numbers. Also, the particle Reynolds number based on the particle-fluid
velocity difference must be smaller than the square root of the particle Reynolds number based
on the shear field. Since this restriction is valid for submicron particles, it is recommended to use
this option only for submicron particles. This study incorporates larger size particle study

Volume 3 Number 4 2011


228 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows
ranging from 50 m to 500 m. And so the Saffman lift forces were not included in the modeling.
The drag force on the particle per unit mass is given in dimensional form as

  18 L C D Re p  
(
Fdrag = FD u u p = )
p d p2 24
(
u up ) (5)

where, rp and dp are particle density and diameter, respectively. Fdrag is non-dimensionalized by
U20 /R. The drag coefficient, CD, is defined as [29]

a2 a
C D = a1 + + 3 (6)
Re p Re 2p

where a1, a2, a3 are the constants, and the particle Reynolds number, ReP, is given as

up u
Re p = d p (7)

Particles are treated as spheres of definite diameter in the above equations.


The particle trajectory known from the particle velocity is given as

dx dy
= up , = vp , (8)
dt dt

where, up and vp are the velocity components of particulate velocity vector, u p . In FLUENT 6.1,
particle trajectory equation, Eq. (4) is integrated using the Trapezoidal scheme. An appropriate time
step is chosen to carry out the trajectory calculations.
The flow domain is the two-dimensional region between the impeller exit and the casing
wall as shown in Fig. 2. The flow rate and average concentration of particles at the impeller
exit are assumed to be prescribed. Particles are assumed to be statistically injected at several
locations on the impeller exit (flow domain entry) with a known mass flow rate of particles.
Particles are assumed to enter into the flow domain with inlet velocity equal to carrier phase
inlet velocity.

Casing Wall
Standard w all func tio ns
Coeff . of restitution

Casing inlet
up = u

Exit (discharge) Zero gradient


u/ dx= v/y= 0

Figure 2. Boundary conditions for the Lagrangian Particle Tracking Method utilized for
dilute slurry flows inside the 2D pump casing domain.

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 229

The boundary conditions are applied at inlet of the casing, exit and along casing walls. At
impeller exit (which is the casing inlet), the solid velocity equals liquid (carrier phase) velocity.
Thus up  u. Along the exit we have zero gradient boundary conditions for the carrier phase as

u v
= =0 (9)
x x

Along the casing walls, coefficients of restitution are used for the solid phase in the normal and
tangential direction that account in their momentum loss. Their mathematical form can be
expressed as Vnnew = en Vnold , Vt new = et Vt old , where en and et are the coefficients of restitution in
the normal and tangential direction. Values of en  1 and et  1 was assumed (elastic collision).
This is justified as the residing time of a particle in the casing is very less. The boundary conditions
as applied to casing domain in dilute slurry flow are illustrated in Fig. 2.
Along the impeller periphery (r  R), the tangential (uq) and radial (ur) velocities for the carrier
phase and the particulate phase are obtained from the specified head (H) and flow rate (Q) as

gH Q
u = and ur = , (10)
U0 2Rb

where b is the impeller width at the outlet, R is the radius of the impeller and h is the hydraulic
efficiency.
Thus the governing equations for dilute slurry flow with Lagrangian-Eulerian approach using
discrete phase model (DPM) in FLUENT 6.1 are described mathematically with two-way
coupling. In the solution to the DPM for dilute slurry flow, SIMPLE algorithm has been applied for
solving the governing equations using the segregated solver. First order upwinding has been used
for discretization. Standard k e model has been used for modeling turbulence.
The trajectory equation of the particle is solved at a frequency of one in ten iterations of the
carrier phase flow equations. The discrete phase and the carrier phase equations are converged to
satisfy a residual convergence of 106 for flow equations with first order upwinding. SIMPLE
algorithm for pressure-velocity coupling is used with first order upwinding in momentum,
turbulence kinetic energy and turbulence dissipation rate terms.

3. RESULTS AND DISCUSSION


The validation of the present results has been carried out in two phases. In the first phase, results
are validated against mesh independence. In the second phase, particle trajectories predicted from
DPM module in FLUENT 6.1 are compared with channel flow results [30] under similar
parametric conditions. Validation of particle trajectories in channel serves as a useful benchmark to
apply DPM module in the pump casing since the actual trajectories of the particles in the pump
casing have not been reported in the literature.
The pump casing considered for presenting mesh refinement studies has a diameter of 0.4074 m
(Casing 1 in Table 1). The flow rate at the inlet is 0.02 m3/s and the corresponding rotation rate is
710 RPM. Three meshes with increasing number of cells in the computational domain are studied
to verify the mesh independent nature of the solutions. The mesh has been graded near the casing

Table 1: Pump casing input data used in the present study at 100% BEPQ (Best Efficiency
Point flow rate)

Casing No. Impeller diameter (m) H (m) Q (m3/s) h RPM


1 0.4074 12.2 0.02 0.73 710
2 0.225 63.5 0.0444 0.71 3400
3 0.395 74 0.088 0.75 1900
4 0.502 77.5 0.17 0.82 1400

Volume 3 Number 4 2011


230 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows
Table 2: Computation time for three meshes

Number of cells Residual convergence Computation time


(in minutes)
18,440 1010 45
30,000 1010 90
40,500 1010 150

C E
0.075 A B

G F

0.05 Computational domain


CS

18,440 cells
30,000 cells
0.025 40,500 cells

0 CS,average (delived concentration)=0.18%


=2650kg/m3
S
A B C D E F G
Points along the casing wall

Figure 3. Mesh independence study: Particle concentration along the casing wall for
three meshes (Casing No. 1 in Table 1).

wall and at the impeller exit region. The computational characteristics of these three meshes in
terms of number of cells in the domain and time of convergence are given below in Table 2. The
densest mesh is found to be three times computationally expensive compared to the coarsest mesh.
In Fig. 3, the comparisons of particle concentrations along the casing wall at various points to
verify the mesh independent nature of the solutions is presented. The inlet particle concentration
is 0.5% (by volume) and the resultant distribution of particle concentration (CS) is also
considered in volume fraction units. A particle with 100 m diameter and a density (rp)  2650
kg/m3 has been considered. The predictions with three different meshes match closely. The
coarsest mesh with 18,440 cells has been finally chosen for the computation of discrete phase
flow. Also seen in Fig. 3, the particle concentration pattern around the shell length shows that the
particle concentration becomes maximum between points F to G and is minimum in the region
from A to B.
The physical quantities such as particle velocity and particle concentration are considered just
in the cell adjacent to the casing wall. It should be noticed that inter-node distance just adjacent to
the casing wall is of the order of 990 m and hence it is sufficient to capture the particle properties
in the cell adjacent to the wall even for a particle diameter of 500 m or a little more. This is shown
in Fig. 4. If the inter node to node distance near the casing wall falls short of the diameter of the
particle then the actual cell should be located where the particle center lies to obtain the accurate
properties. In our simulations, the particle diameter completely lies inside the very cell adjacent to
the casing wall.

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 231

Figure 4. Meshed domain with node to node distance near the casing wall. Note the
node to node distance is 990 microns near the casing wall.

In second phase of the validation, comparison of particle trajectories predicted by FLUENT 6.1
with numerical data published in the literature is carried out for the channel flow problem since
actual particle trajectories are not available for pump casing in the literature. In Fig. 5, comparison
between present predictions and Tsuji et al. [30] has been made for particle trajectories inside a
channel. The channel height is normalized to unity. An aspect ratio of 1:200 (height to length) is
considered. Three particles are released at 14, 12 and 34 height of the channel. The dashed line denotes
the present predictions and the solid lines denote the results from Tsuji (1987). The predictions
between the two numerical solutions compare favorably both qualitatively and quantitatively
except a small lateral shift in the trajectories.
Four pump casings are considered for modeling in the present work. Their sizes are different with
different duty conditions. The four casing geometries as described briefly in Table 1 have their
diameters as 0.4047 m, 0.225 m, 0.395 m and 0.502 m. The duty conditions are provided for
(maximum) Best efficiency point flow rate (BEPQ) at a given RPM. The input data set contains

Figure 5. Comparison of present predictions of particle trajectory inside a


channel with Tsuji et al. [30].

Volume 3 Number 4 2011


232 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows
Table 3. Pump casing input data for Casing No.3 in Table 1.

Flow Rate in m3/hr


(m3/s) 240 320 400 480
(.0666) (.0888) (.1111) (.1322)
RPM
(60% BEPQ) (80% BEPQ) (100% BEPQ) (120% BEPQ)
U0 (Tip Velocity)
1500 H  47m H  41m H  35m n. a.
U0  31.01 h  0.76 h  0.76 h  0.70
uq  19.54 uq  17.05 uq  15.80
ur  0.599 ur  0.796 ur  1
1700 H  62.5m H  56.5m H  49m H  42m
U0  35.14 h  0.75 h  0.77 h  0.75 h  0.7
uq  23.2 uq  20.46 uq  18.22 uq  16.73
ur  0.599 ur  0.796 ur  1 ur  1.165
1900 H  81m H  74m H  66m H  57m
U0  39.28 h  0.73 h  0.77 h  0.77 h  0.74
uq  27.68 uq  23.98 uq  21.38 uq  19.22
ur  0.599 ur  0.796 ur  1 ur  1.165

Head (H), Flow rate (Q), h (efficiency), and RPM for a given pump casing. The data points for
several other RPM are not provided here, only one RPM has been selected to represent the particular
pump casing. For an ease of understanding, the input set consisting of Head (H), flow rate (Q),
tangential velocity (uq) and radial velocity (ur) are provided for Casing 3 in Table 1 for three RPM
viz. 1500, 1700 and 1900 in Table 3. All entries in the table are in respective dimensional units.
In slurry flow, as aforementioned, erosion is caused by two mechanisms viz. particle impact and
sliding. Since in dilute slurry flows, the particles do not slide along the wall rather they move by
impinging on the wall, the sliding wear mechanism does not hold true for the dilute flows. As
described previously, impact erosion takes place due to particles hitting the casing wall. It was also
described that the specific erosion rate for a particle having a speed VS hitting the wall at an angle
of a [14] is

Wsp = ( ) p CS VSm , m = 3 (11)

where rp (assumed constant) is the density of the solid particles and CS is the particle concentration.
The speed VS of the particle is the velocity (magnitude) of the particle in the cell adjacent to the
casing wall. The expression f(a) takes into account of the particle, slurry and slurry pump casing
material and varies for different combinations and different angles. As the impact angle varies
marginally around the shell perimeter, the variations in the parameter f(a) are bound to be
marginal. Hence, it would be sufficient (to capture the essence) to compare the value of the factor
V 3S CS (hereafter referred to as impact wear parameter) along the casing wall for various
operational and geometry conditions of the pump casing. Hence, in our studies for impact wear rate
prediction, we leave out the parameter f(a) and only compare the parameter V 3S CS along the casing
wall.
The comparison of impact wear parameter has been made for various operating and modified
geometry conditions. Operating conditions include flow rate (expressed in terms of % BEPQ), and
pump speed (RPM). Geometry modified conditions include different tongue curvature, different
casing width, impeller diameter and different discharge pipe slope. The definition of each of these
will be expressed quantitatively when these parameters appear in context.

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 233

C E

A B

G F

VS 0.3

0.2

0.1
A B C D E F G
Points A long the Casing Wall

Figure 6. Particle velocity along the casing wall with sharp discontinuity
near the tongue region.

The impact wear rate parameter (V 3S CS) contains the particle velocity and particle concentration
along the casing wall. Before presenting the impact wear rate parameter variations, it is intended to
present these quantities separately. The representative trends of particle velocity and particle
concentration along the casing wall are shown in Fig. 6 and Fig. 7, respectively.

C E
0.9 B
A

0.8 G F

0.7
Inlet CS = 0.5%
0.6

0.5
CS

0.4

0.3

0.2

0.1

0
A B C D E F G
Points Along the Casing Wall

Figure 7. Particle concentration along the casing wall at various points (A, B, C, D, E, F
and G) along the casing wall for Casing No. 3 in Table 1.

Volume 3 Number 4 2011


234 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows

Figure 8. Particle (100 m diameter) trajectories inside the pump casing showing that
the particles eventually strike the casing wall.

Note the trends of variation in velocities and concentration of particles along the casing wall. It
is observed that the velocity increases along the casing wall from point B to F. A sharp discontinuity
occurs in the velocity in the tongue region (Point B). Similar to the velocity along the casing wall,
the concentration along the casing wall also increases monotonically from B to F.
The particle trajectories eventually strike the casing wall as seen in Fig. 8 and thus the
concentration tends to increase as we move more downstream in the belly region of the pump
casing.
The impact wear parameter given as V 3S CS represents the impact wear rate along the casing wall.
Its variation is non-uniform along the casing wall and it monotonically increases from the tongue
region to the belly region of the casing. Reduction in the peak value of the impact wear parameter
and making it more uniform by applying several operational and geometry modifications is
presented in the following section.
In the following, the effect of various parameters viz. flow rate, pump speed (RPM), tongue
curvature, casing width, impeller diameter and particle diameter on impact wear parameter is
discussed.

3.1 Effect of Flow Rate


The impact wear parameter (V 3S CS) varies with different flow rates along the casing wall. The
trends of variation can be seen in Fig. 9 where with reduced flow rate, it is found that the magnitude
of the impact wear parameter reduces and it tends to even out along the casing wall. The impact
wear parameter has been presented in normalized form with the normalizing value taken as the
impact wear parameter for 120% BEPQ (Best Efficiency Point Flow Rate). As observed, 120%
BEPQ has the maximum peak value and 60% BEPQ has the minimum value. It is attributed to the
fact that with decrease in the flow rate, particle concentration and velocities tend to decrease near
the casing wall. Velocities near the wall should be compared to substantiate this fact. The trends of
velocities are seen in Fig. 10 for two flow rates viz. 100% BEPQ and 60% BEPQ. For the higher
flow rate (here 100% BEPQ), the velocities near the wall tends to increase but for lower flow rate
(here 60% BEPQ), the velocities near the wall tends to decrease along the shell length from point
B to F. Particularly near the point F on the casing wall, velocities for the higher flow rate are higher
than the velocities for the lower flow rates. Another fact is that at lower flow rates (lower % BEPQ),
the pump tends to recirculate more amount of liquid around the impeller as the tangential velocity
(uq) tends to be higher and radial velocity (ur) tends to be lower. This can be appreciated from the
expressions of the velocities at the impeller exit given as

gH Q
u = ; ur = (12)
U0 2Rb

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 235

D
Normalization is with respect to V3SCS
for Q=120% BEPQ
C E
1
A B
0.9 G F

0.8

PQ
0.7

BE
0%
0.6

12
V 3SC S
0.5 PQ
BE
%
0.4 100
PQ
BE
0.3 %
80
0.2
P Q
0.1 60% BE

0
A B C D E F G
Points Along the Casing Wall

Figure 9. Comparison of impact wear rate parameter (V 3S CS) with different flow rates viz.
120% BEPQ, 100% BEPQ, 80% BEPQ and 60% BEPQ. Note that 100% BEPQ
corresponds to a flow rate of 0.11 m3/s (Refer Casing No. 3 in Table 1).

From the typical drooping Q-H curve (for a given RPM), as Q decreases, H increases. Thus with
lower Q, uq increases and ur decreases.
Similarly, concentration for two flow rates viz. 60% BEPQ and 100% BEPQ is compared in Fig.
11. It is seen from the figure that the particle concentration for the 100% BEPQ flow rate is higher
in all regions of the pump casing compared to 60% BEPQ except near the point F.
It is seen that the flow rates at 80% or below produce lesser impact wear along the casing wall
which tends to even out as the flow rate further decreases.
D

C E
0.6
A B

G F

0.5

0.4
VS

EPQ
1 00 % B
0.3

60% B E
PQ
0.2

A B C D E F G
Points Along the Casing Wall

Figure 10. Comparison of particle velocities near casing wall for 60% BEPQ (0.066
m3/s) and 100% BEPQ (0.11 m3/s) for Casing No. 3 in Table 1.

Volume 3 Number 4 2011


236 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows
D

C E

A B
0.4
G F

0.3

PQ
CS

0.2 BE
0%
10
PQ
BE
%
0.1 60

0
A B C D E F G
Points Along the Casing Wall

Figure 11. Solid particle concentration along the casing wall for 60%
BEPQ and 100% BEPQ.

3.2 Effect of Pump Speed


Three pump speeds viz. 1500 RPM, 1700 RPM and 1900 RPM are considered. At the lowest RPM,
the impact wear parameter is found to be minimum and at the highest RPM viz. 1900 RPM, the
normalized impact wear parameter peak value is maximum at 1 compared to a maximum of 0.76
at 1500 RPM (see Fig. 12).
The cases run correspond to pump Casing No. 3 in Table 1. For 1900 RPM, a flow rate of 0.11
m3 is considered. The corresponding points on other RPM (1500 and 1700) are determined using
the pump scaling laws given as
2
H1 N1 Q N
= and 1 = 1 (13)
H2 N 2 Q2 N2

where N represents pump speed (in RPM).


Higher impact wear parameter at higher RPM is quite obvious since the flow velocities are
increased. Practical experience [23] also confirm these trends.

3.3 Effect of Tongue Curvature


In this section, we consider the effect of one of the geometrical parameter viz. tongue curvature.
This has been expressed mathematically by defining a ratio (length OB)/R (length OB to Radius
ratio) as shown in Fig. 13. For a normal pump casing with original geometry, the OB/R ratio equals
1.25. A slight modification is done in this ratio and is changed to 1.3 with point B taken outwards
and 1.2 with point B taken slightly inwards. The location of point B is very critical since it controls
the flow velocity direction and magnitude as well.
Figure 14 shows the respective trends of three tongue curvature geometries with OB/R ratio
equal to 1.2, 1.25 and 1.3. With the tongue slightly drawn inwards, we find a drastic reduction in
the impact wear parameter along the casing wall. Whereas for the tongue drawn outwards, we find
an increase rather than a decrease in the impact wear parameter.
Velocities near the wall vary significantly for the different tongue geometries with different
OB/R ratio. Fig. 15 shows that velocities for OB/R  1.3 geometry are greater than OB/R  1.2
type of tongue geometry. Surprising results are seen in relation to trends of velocity along the
casing wall. Firstly, the flow is not only changed locally near the tongue but all along the wall.
Secondly, the velocities as seen along the casing wall are expected to decrease in general with
tongue drawn outwards and not increase. The trends however contradict such a priori intuition.

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 237

1.0 C E

0.9 A
B
0.8
Normalized V 3SC S
G F
0.7 1500 RPM

0.6
1700 RPM
0.5
0.4 1900 RPM
0.3
0.2
0.1 Normalized with the peak value of
impact wear parameter of 1900 RPM
0.0
A B C D E F G
Points along the casing wall
Figure 12. Variation of impact wear parameter along the casing wall for different pump
speed for Casing No. 3 in Table 1.

Tongue region D

R
C O E

A
B

G F OB/R=1.25

Figure 13. OB to R ratio shown as criterion for defining tongue curvature.

In Fig. 15, only a local increase in the velocity is seen in the tongue region when it is drawn
inwards but away from the tongue the velocities happen to be lower compared to geometry with
tongue drawn outwards. This contradicts the intuitive understanding of the flow. Thus CFD
simulations give surprising results which are not comprehended using an intuitive sense.
Also evident from Fig. 16, the concentration for the OB/R  1.3 value exceeds the concentration
of tongue with OB/R  1.2. Thus the cumulative effect of reduced velocity and concentration leads
to drastic reduction in the impact wear parameter for OB/R  1.2 as evident from Fig. 14.

3.4 Effect of Discharge Slope


The peak of the impact wear parameter occurs near the point F on the casing wall. This may be reduced
with increased slope of the discharge pipe. The effect of this is illustrated in Fig. 17. By slanting the
discharge, the peak value of the impact wear parameter has reduced slightly. But near the casing exit
(point G), the amount of reduction is quite high with the slanted discharge. Since the particles tend to
sweep away and do not accumulate on account of slanted length of the discharge pipe, the slanted

Volume 3 Number 4 2011


238 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows
OB/R=1.2 OB/R=1.25 OB/R=1.3

C E
1 A B

0.9 G F

0.8
0.7
OB/R=1.2
0.6 OB/R=1.25
V 3SC S

OB/R=1.3
0.5
0.4
0.3
0.2
3
0.1 Normalization is with respect to VSCS
for OB/R =1.3
0
A B C D E F G
Points Along the Casing Wall

Figure 14. Trends of impact wear parameter (V 3S CS) for different tongue geometries viz.
OB/R  1.2, OB/R  1.25 and OB/R  1.3 in Casing No. 3 in Table 1.

OB/R=1.2 OB/R=1.25 OB/R=1.3 Velocity near point B


0.4

C E
0.3

A B

0.2
G F OB/R=1.2
OB/R =1.25
OB/R=1.3

0.4

0.3
VS

0.2
OB/R=1.2
OB/R =1.25
OB/R=1.3
0.1

A B C D E F G
Points Along the Casing Wall

Figure 15. Velocities along the casing wall for OB/R  1.2, OB/R  1.25 and OB/R 
1.3 shown as a supporting figure to Fig. 14.

discharge has lower values of impact wear rate parameter. Numerical experiment indicate (not shown
here) a greater slope in the slanted length will reduce the peak impact wear rate to a greater level.

3.5 Effect of Casing Width


Another important parameter that can affect the impact wear parameter significantly along the
casing wall is the casing width b. It is important to consider again the form of the velocity at the
inlet as given by

gH Q
u = ; ur = (14)
U0 2Rb

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 239

OB/R=1.2 OB/R=1.25 OB/R=1.3

C E

0.8 A B

G F
0.7

0.6 OB/R=1.2
OB/R=1.25
0.5 OB/R =1.3
CS

0.4

0.3

0.2

0.1

0
A B C D E F G
Points Along the Casing Wall

Figure 16. Concentration along the casing wall for OB/R  1.2,
OB/R  1.25 and OB/R  1.3.

C E
1 A B

F SLANTED DISCHARGE
0.9 G STRAIGHT DISCHARGE

0.8
straight discharge
0.7 Slanted discharge

0.6
V 3SC S

0.5
0.4
0.3
0.2
0.1
Normalization is with respect to V 3SCS for straight discharge
0
A B C D E F G
Points Along the Casing Wall

Figure 17. Impact wear parameter (V 3S CS) along the casing wall for straight
(point G lies 305 mm below origin in Y direction) and slanted discharge
(point G 315 mm from the center in Y axis).

Here, b is the casing width at the outlet of the impeller. As it is made large, we notice a decrease in
the radial velocity. The results corresponding to different casing widths are shown in Fig. 18.
As seen, there is a reduction of impact wear parameter with increased casing width. Four casing
widths viz. b  0.8b0, b  b0, b  1.2b0 and b  1.3b0 are considered for the same pump casing.
The pump casing with a minimum casing width shows the maximum value of the impact parameter
and the pump casing with the maximum casing width shows the minimum impact parameter along
the casing wall.

Volume 3 Number 4 2011


240 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows
D 3
Normalization is with respect to VSCS for b=0.8b0

C E

A B
0.9 F
G

0.8
0.7 b=0.8b0
b=b0

0
.8b
0.6 b=1.2b0

b =0
b=1.3b0
V 3SC S

0.5
b=b 0
0.4 . 2b 0
b=1
0.3 . 3b 0
b=1
0.2
0.1
0
A B C D E F G
Points Along the Casing Wall

Figure 18. Variation of impact wear parameter along the casing wall for various casing
widths viz. b  0.8b0, b  b0, b  1.2b0 and b  1.3b0.

3.6 Effect of Impeller Diameter


Another important geometrical parameter that can affect the impact parameter significantly is the
impeller diameter. Three impeller diameters are studied viz. d  0.95d0, d  d0, and d  1.05d0.
The head and flow rate for the given pump are modified for the changed impeller diameter
following [31] as

3 2
Q1 D H D
= 1 and 1 = 1 (15)
Q2 D2 H2 D2

The results are shown in Fig. 19 with the impact wear parameter plotted along the casing wall. As
the impeller diameter decreases from d  1.05d0 to d  0.95d0, the impact wear rate parameter is
reduced from a normalized value of 1 to 0.63.

3.7 Effect of Particle Size


Impact wear rate parameter also depends on the size of the particle. Four mono-size slurries with
particle sizes of 100 m, 200 m, 500 m and 1000 m, respectively, are studied to determine the
effect on impact wear parameter along the casing wall in Fig. 20. It is observed that with increased
diameter of the particle, the impact wear parameter increases. In Fig. 21 the particle trajectories of
these four particles are compared. As seen in the figure the largest diameter particle reaches the
casing wall soonest. This tends to increase the particle concentration of the larger diameter particles
to be higher than smaller diameter particles near the casing wall.
As the particle size increases, the particle tends to get through the exit very quickly and
recirculation around the impeller tends to be low. The particle trajectories for a 100 m and 500 m
are shown in Fig. 22. It is reflected in the figure that the 100 m particle tend to recirculate more
past the impeller before it exits the discharge.
So far it has been noticed that the effect of the tongue curvature drawn slightly inwards, lower
flow rate, lower RPM, slanted discharge, higher casing width and smaller impeller diameter reduce

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 241

d=0.95d0
Normalization is with respect to V3SCS of d=1.05 d0

1.0

0.9

0
d
d=d0 D
0.8

5%
10
0.7 C E

a=
A B

Di
0.6
G F
V 3SC S

0.5 d=1.05d0

0
0.4

=d
0.3
D ia
0

d
0.2
%
95
=

0.1
ia
D

0.0
A B C D E F G
Points Along the Casing Wall

Figure 19. Impact wear parameter along the casing wall for various impeller
diameter viz. d  0.95d0, d  0 and d  1.05d0 with d0  395 mm
corresponding to Casing No. 3 in Table 1.

D
3
Normalization is with respect to VSCS
for particle with dia = 1000 m
C E
1
A B

0.9 F
G

0.8

0.7
100 m
0.6
200 m
V 3SC S

0.5
500 m
0.4
1000 m
0.3

0.2

0.1

0
A B C D E F G
Points along the Casing Wall

Figure 20. Impact wear parameter (V 3S CS) along the casing wall for various particle
diameters of 100 m, 200 m, 500 m and 1000 m.

Volume 3 Number 4 2011


242 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows

1 10
00
m
0.5

0
Y

-0.5

m
0
50
-1
100 m
200 m
-1.5
-2 -1 0 1
X

Figure 21. Particle trajectories inside the casing wall for various particle diameters of
100 m, 200 m, 500 m and 1000 m in Casing No. 3 in Table.1.

Figure 22. Recirculation past the impeller for 100 micron and 500 micron sized particles.

the impact wear parameter along the casing wall and tend to even it out along the shell. Some of
these parameters may be coupled together to consider the effect on impact wear parameter.
Three different cases are tried on the same pump casing (Casing No. 3 in Table 1). In the first
case (CASE A), normal flow rate (100% BEPQ) and the normal designed geometry (OB/R  1.25)
is considered. In the second case (CASE B), an OB/R  1.2 has been provided with increased
casing width i.e. b  1.25b0 and finally in the third case (CASE C), change the flow rate from 100%
BEPQ to 80% BEPQ is considered.
For Case C, not only a significant reduction in the impact wear parameter along the casing wall
but also very close to an even distribution (of impact wear parameter) along the casing wall is
observed. While for the normal flow and geometry (Case A), the impact wear parameter tends to
be very uneven along the shell and the peak also tends to be very high. Compared to the normal
flow and geometry conditions in Case A, the impact wear parameter in Case C is just 10% of the
CASE A value as seen in Fig. 23.
At certain operating conditions, reverse flow may also become prominent near the tongue
region. In particular this takes place at higher flow rates, when the radial velocity is high. Fig. 24
shows reverse flow taking place near the tongue region (which is also shown enlarged). The

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 243

D
3
Normalization is with respect to VSCS for CASE A
C E
1
A B

0.9 G F

0.8
CASE A
0.7
CASE B
CASE C
0.6
V 3SC S

0.5

0.4

0.3

0.2

0.1

0
A B C D E F G
Points along the Casing Wall

Figure 23. Variation in the impact wear parameter for different combinations
of various parameters: Case A: 100% BEPQ with OB/R  1.25 and b  b0,
Case B: 100% BEPQ with OB/R  1.2 and b  1.25 b0 and Case C: 80%
BEPQ with OB/R  1.2 and b  1.25 b0.

3
Pump Dia. = 502 mm, RPM = 1400, Q = 0.28 m /s, H= 62m

=U0 (1 unit)
1
Reference vector

0.5

0
y

-0.5

-1

-1.5
-1 0 1
x

Figure 24. Velocity vectors showing reverse flow near the tongue for Casing No. 4 in
Table 1 at 100% BEPQ (0.17 m3/s).

particular pump casing studied here has diameter 502 mm (Casing. No. 4 in Table 1). Reverse flow
can affect the wall stresses and the tangential velocity near the tongue significantly.
The effect of various geometry and operational parameters on the variation of impact wear
parameter inside the pump casing is discussed. The description aims to reduce the impact wear
parameter along the casing wall and to even it out as far as possible.

Volume 3 Number 4 2011


244 CFD Prediction of Erosion Wear in Centrifugal Slurry Pumps for
Dilute Slurry Flows
4. CONCLUSION
Mesh independence of the solutions is established for particle concentration inside the pump
casing. Validation of present predictions is carried out for the 2D channel flow problem with
published numerical data.
Effect of several operational parameters viz. pump flow rate, pump speed (RPM), particle
diameter and various geometry conditions viz. tongue curvature, slope of the discharge pipe and
casing width is studied. Rather than incorporating energy parameter which relates the amount of
energy that a particle loses with the erosion rate, the impact wear parameter V 3S CS is calculated for
a physical and qualitative representation of the impact wear rate. In all these studies, the effect is
evaluated in terms of the impact wear parameter.
Erosion prediction is made for different operational and geometric parameters with an idea to
lower the wear rates and to make the wear pattern along the casing wall as uniform as possible. The
operational parameters constitute pump flow rate (measured in terms of % BEPQ). The most
important finding is that as the flow rate decreases from 120% BEPQ to 60% BEPQ, the peak
impact wear parameter shifts from the belly region (point F) to the tongue region (point B). For the
cases studied at 80% BEPQ, the peak impact wear parameter is found to be minimum and the wear
pattern is nearly uniform. Another important operational parameter is pump speed (RPM). With a
decrease of RPM, the wear rate subsides.
In geometric parameters, tongue curvature, slope of the discharge pipe and casing width are
studied. The tongue curvature has been mathematically described as a ratio of two lengths (OB and
R). With the tongue drawn slightly inside, a great reduction in the impact wear parameter is seen and
with the tongue drawn slightly outwards, an increase in the wear rate is seen. Similarly with increase
in the casing width, decrease in the impact wear parameter is seen and with a decrease of the casing
width, an increase in the wear parameter is seen. The peak impact wear parameter occurring near the
belly region of the pump casing at point F is treated also using a slanted discharge over a straight
discharge.

ACKNOWLEDGEMENTS
The work has been carried out at the Indian Institute of Technology Delhi, New Delhi as part of the
dissertation. The authors gratefully acknowledge the computational facilities extended by the
Department of Applied Mechanics for carrying out the work.

REFERENCES
[1] K.C. Wilson, G.R. Addie and R. Clift, Slurry Transport Using Centrifugal Pumps, Elsevier Science Publishers Ltd.,
Essex, England, 1992.
[2] A. Sellgren and L. Vappling, Effects of Highly Concentrated Slurries on the Performance of Centrifugal Pumps, Int.
Symposium on Liquid-Solid Flows, New York, ASME-FED, Vol.38, 47pp. 143148, eds. M.C. Roco, and W.
Weidenroth, 1986.
[3] Sellgren A., Addie G.R., Visintainer R, Pagalthivarthi K. 2005: Prediction of slurry pump component wear and
cost, Proceedings, WEDA XXV and Texas A& M Annual Dredging Seminar, New Orleans, U.S.A., June.
[4] G. R. Addie, A. Sellgren, J. Mudge, SAG mill pumping cost considerations. SAG Conference, 3rd International
Conference on Autogenous & Semiautogenous Grinding Technology, September 30October 3, 2001, Vancouver,
B.C., Canada, 2001.
[5] Roco, M.C., Nair, P., Addie, G.R. and Dennis, J., 1984, Erosion of Concentrated Slurry in Turbulent Flow, ASME
FED, Vol. 13, pp. 6977.
[6] Roco M C and Cader T 1988, Numerical method to predict wear distribution in slurry pipelines. Advances in Pipe
Protection. BHRA Fluid Eng., Cranfield, UK, pp 5385.
[7] Addie, G.R. and Pagalthivarthi, K.V., 1989, Prediction of Dredge pump shell wear, Proc. WODCON XII, 12th
world Dredging Conference, pp. 481504.
[8] Minemura, K. and Uchiyama, T., 1990, Calculation of the three-dimensional behavior of spherical solid particles
entrained in a radial-flow impeller pump, Proc. Instn Mech Engrs., Vol.204, pp.159168.
[9] Pagalthivarthi, K.V. and Helmly, F.W., 1992, Applications of Materials Wear Testing to Solids Transport via Centrifugal
Slurry Pumps, Wear Testing of Advanced Materials, ASTM STP 1167, eds. Divakar, R. and Blau, P.J., pp. 114126.
[10] Tuzson, J.J. and Scheibe-Powell, K.A., 1984, Slurry Erosion Tests with Centrifugal Erosion Tester, Liquid-Solid
Flows and Erosion Wear in Industrial Equipment, ASME-FED, Vol.13, ed. Roco, M.C., pp. 8487.

Journal of Computational Multiphase Flows


K. V. Pagalthivarthi, P. K. Gupta, Vipin Tyagi, and M. R. Ravi 245

[11] H.H. Tian, G.R. Addie, K.V. Pagalthivarthi, Determination of wear coefficients for prediction through Coriolis wear
testing, WEAR, Vol. 259, pp. 160170, 2005.
[12] H.Mc.I.Clark, J. Tuzson and K.K. Wong, Measurements of specific energies for erosive wear using a Coriolis erosion
tester. Wear 241 (2000), pp. 19.
[13] K. V. Pagalthivarthi and G. R. Addie, Prediction methodology for two-phase flow and erosion wear in slurry
impellers, 4th International Conference on Multiphase Flow, ICMF-2001, New Orleans, LA, May 27-June 1, 2001.
[14] Roco, M.C. and Addie, G.R., 1983, Analytical Model and Experimental studies on Slurry flow and erosion in pump
casings, Slurry transportation, STA, Vol.8, pp.263.
[15] Roco, M.C., Addie, G.R. and Visintainer, R.J., 1985, Study of Casing Performances in Centrifugal Slurry Pumps,
Particulate Science and Technology, No.3, pp. 6588.
[16] Pagalthivarthi, K.V., Desai, P.V. and Addie, G.R., 1990, Particulate Motion and Concentration fields in Centrifugal
Pumps, Particulate Science and Technology, No.8, pp. 7796.
[17] Zhong, Y. and Minemura, K., 1996, Measurement of erosion due to particle impingement and numerical prediction
of wear in pump casing, Wear, Vol. 199, pp. 3644.
[18] Bitter, J.G.A., A study of Erosion Phenomenon, 1963, Wear, Vol. No. 6, pp. 521.
[19] Addie, G.R., Pagalthivarthi, K.V. and Kadambi, J.R., 2004, PIV and finite element comparisons of particles inside
a slurry pump casing, Proc. Int. Conf. on Hydrotransport 16, Santiago, Chile, pp. 547559.
[20] Charoenngam, P., Subramanian, A., Kadambi, J. R., Addie, G. R., 2001, Investigations of slurry flow in a
centrifugal pump using particle image velocimetry, 4th International conference on multiphase flow, New Orleans,
May 27-June 1.
[21] M. Asuaje, F. Bakir, S. Kouidri, F. Kenyery, R. Rey, Numerical modelization of the flow in centrifugal pump: volute
influence in velocity and pressure fields, International journal of rotating machinery, 2005, Vol. 3, pp. 244255.
[22] Visintainer, R.J., Addie, G.R. and Pagalthivarthi, K.V., 1992, Prediction of Centrifugal Slurry Pump Wear,
International Conf. on Pump and system, Beijing, China, May 1921.
[23] Addie, G.R., Pagalthivarthi, K.V. and Visintainer, R.J., 1996, Centrifugal Slurry Pump Wear, Technology and Field
Experience, ASME Fluids Meeting, San Diego.
[24] J.S. Shirolkar, C.F.M. Coimbra and M.Q. McQuay, Fundamental Aspects of Modeling Turbulent Particle Dispersion
in Dilute Flows, Prog. Energy Combustion Science, Vol.22, pp. 363399, 1996.
[25] Lu, Q.Q., Fontaine, J.R. and Aubertin, G., 1993, A Lagrangian model for solid particles in turbulent flows, Int J
Multiphase Flow, Vol No.19 (2), pp. 347367.
[26] Wang, J., Shirazi, SA., Shadley, J.R. and Rybicki, E.F., 1996, Application of flow modeling and particle tracking to
predict sand erosion rates in 90-degree elbows FED-Vol. No. 236, ASME fluids engineering division conference,
Vol.1, pp. 725734.
[27] Keating, A. and Nesic, S., 2000, Particle tracking and erosion prediction in three-dimensional bends, Proc. 2000
ASME fluids engineering summer meeting, Pap. No. FEDSM2000-11249.
[28] Wallace, M.S., Peter, J.S., Scanlon, T.J., Dempster, W.M., McCulloch, S. and Ogilvie, J.B., 2000, CFD-based
erosion modeling of multi-orifice choke valves, Proc. 2000 ASME fluids engineering, Boston, Pap. No. FEDSM
200011244.
[29] S.A. Morsi and A.J. Alexander, An Investigation of Particle Trajectories in Two-Phase Flow Systems, J. Fluid Mech.
Vol. 55(2), pp. 193208, 1972.
[30] Y. Tsuji, Y.Morikawa, T. Tanaka, N. Nakatsukasa and M. Nakatani, Numerical Simulation of Gas-Solid Two-Phase
Flow in a Two-Dimensional Horizontal Channel, 1987.
[31] Karassik, E., Krutzsch, W.C., Fraser, W.H. and Messina, J.P., 1985, Pump Handbook, 2nd Edition, Mc-Graw Hill.

Volume 3 Number 4 2011

Вам также может понравиться